Longer video clips and editing through conversation
Google released Gemini Omni 1.2 Flash on November 5, 2026. The model makes short videos from prompts and reference media and edits existing footage with written instructions. The release raises the maximum clip length from 10 to 15 seconds, improves how the model follows references according to Google, and brings it to general availability in the Gemini API and on Google Cloud.
Make a clip from prompts and references
A clip can begin with a written prompt, a still image, or start and end frames. References can guide different parts of a scene. A character image can guide appearance, a location image can establish the setting, and a short video can guide movement. Google’s launch examples use motion references of up to three seconds.
Omni can generate sound with the video. Google demonstrated dialogue with lip-sync and separate voice references for multiple speakers. Voice-reference features vary by product and region. On Google Cloud, access to audio references requires approval.
Revise a scene in conversation
A follow-up prompt can ask Omni to change the weather, replace a prop, or switch the background. For example, a clip of a cyclist crossing a bridge could be edited to show rain while asking the model to keep the cyclist’s movement and the camera path. Flow’s editing history retains previous versions, and its refine workflow supports up to three conversational turns with earlier edits in context.
Continue beyond the first shot
On the web version of Flow, Omni Extend can add three to ten seconds to a clip on Google AI Plus, Pro, or Ultra. Up to three reference images can guide the continuation. The same launch retired Veo Extend in Flow.
Availability and production limits
The 15-second limit and editing options depend on where the model is used.
Product | Access at launch |
|---|---|
Google Flow | Core generation and editing modes offer 3–10 seconds without a subscription and 3–15 seconds with Google AI Plus, Pro, or Ultra. Start and end frames are a web feature. |
Gemini app | Video generation requires a Google AI plan or a qualifying Workspace license. The 15-second option is rolling out to Ultra. Other eligible users have a 10-second limit. |
Google Vids | Google announced up to three 15-second generations per month at no cost with a Google account or Workspace subscription. |
Gemini API and Google Cloud | Generally available. Cloud documentation lists 3–15 second outputs at 360p, 720p, 1080p or 4K, with 16:9 and 9:16 aspect ratios for text-to-video generation. |
Pricing and resolution
Google listed the 720p API price at US$0.07 per second, 30% below Omni 1.1 Flash. At that rate, a 15-second clip costs US$1.05. Flow uses credits, with costs that vary by duration and resolution. Its 360p drafts can be upscaled to 1080p on eligible plans or to 4K with Ultra.
Consistency and regional access
Google’s own before-and-after examples show that small visual details can still change. The company identifies fast sports action, rapid script changes, and complex physics as areas for improvement. Editing uploaded footage and using certain voice features also depend on regional availability.
Sources
Google launch announcement · Google DeepMind model card
Flow model guide · Flow editing guide · Gemini app help
Google Vids launch · Google Cloud launch and pricing · Cloud output specifications